INDEX

Explanations

network-related terms and entity names, such as addresses, ports, editing, software, and security settings

oai_token-act-pair · gpt-3.5-turbo

conjunctions and phrases that indicate relationships between technical components or conditions

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

itans

-0.74

uta

-0.74

bo

-0.74

aws

-0.69

poke

-0.68

Shut

-0.65

fur

-0.64

Ĥİ

-0.64

odcast

-0.63

going

-0.63

POSITIVE LOGITS

 hence

1.23

 thereby

1.21

 consequently

1.20

 thus

1.14

 therefore

1.13

 optionally

1.07

 thence

0.95

rogens

0.93

 possibly

0.91

 subsequently

0.89

Activations Density 0.952%

network-related terms and entity names, such as addresses, ports, editing, software, and security settings

conjunctions and phrases that indicate relationships between technical components or conditions

No Comments

No Known Activations